The density of electronics components on a semiconductor chip structure, which has increased one thousand-fold in the last ten years, is limited in part by the need to remove heat generated by operation of these chips. Chips have been moved closer together in order to reduce the time for a signal to propagate from one chip to another. Some components, such as static and dynamic random access memory units ("RAMs") generate relatively little heat in operation, as little as 0.25 Watts per unit. Other semiconductor devices, such as signal processing chips, produce thermal energy at a rate of 5-35 Watts. As the real estate on a semiconductor structure that is allocated to a given chip is reduced, the problem of heat generation becomes more intractable, for at least two reasons: (1) The heat produced per unit area and per unit time increases proportionally with the chip density; and (2) A given chip may generate heat at a much higher rate than its nearest neighbor chips and thus contribute unexpectedly to the thermal environment in which these nearest neighbor chips operate.
Several approaches have been devised to remove heat from a semiconductor device. Mayerhoff et al., in U.S. Pat. No. 3,361,195, disclose use of a heat sink including a plurality of passageways in a substrate, the passageways having an annular or radial, serpentine arrangement and being filled with a heat sink liquid that is force-flowed through the passageways to accept and carry off heat from the device. This approach requires use of an external plenum to accept and cool the returning liquid, requires use of an engine to force the liquid through the passageways, and requires provision of a structure containing the serpentine passageways for each semiconductor device or group of such devices.
In U.S. Pat. No. 3,564,109, Ruechardt discloses use of cooling ribs or fins at the housing wall of a device from which heat is to be removed. The ribs from a single piece and are fabricated from a special heat conducting synthetic material such as a polyester resin, an epoxide resin araldit, silicon or polyproylene. The rib shapes are molded to increase the ratio of exposed cooling area to volume of heat-producing material, and the ribs may have internal ducts through which a liquid is force-flowed.
The heat pipe principle for transport of heat to another location was first discussed by Gaugler in 1944 (U.S. Pat. No. 2,350,349) and by Trefethen in 1962 (G.E. Co. Tech. Info. Ser. No. 615-D114). Grover and co-workers independently invented and fabricated a heat pipe device in 1964 (Jour. of Appl. Phys., vol. 35 (1964) 1190-1191) and demonstrated its effectiveness as an approximately isothermal, heat transmission device.
Transistor cooling by use of a heat pipe with a dielectric powder wick material is disclosed by Sekhon et al. in U.S. Pat. No. 4,047,198. The dielectric wick material is sprayed onto or adhered to all interior surfaces of the heat pipe structure and to all exposed surfaces of the electronic device(s) to be cooled. The wick material is a conformal coating of dielectric powder, which serves as a mat, followed by bundles of glass fibers of average length 0.08 cm and average diameter 0.0008 cm. The heat pipe working fluid is a fluorochemical liquid. The cooling mechanism is not an integral part of the electronic assembly.
Sasaki et al, in U.S. Pat. No. 4,327,399, discloses use of a heat pipe cooling arrangement for IC chips in which the IC is inserted in an aperture in a substrate that directly communicates with a heat pipe cavity containing a wick and a working fluid. The IC also carries a wick, which is aligned with the wick in the heat pipe cavity for direct cooling of the IC. The substrate must have apertures formed therein in order to provide direct access to the working fluid.
Substrate cooling by heat pipe apparatus is disclosed by Wiech in U.S. Pat. No. 4,519,447. The heat pipe contains a working fluid that, in liquid form, passes close to one or more heat-producing components that are positioned on the substrate. A working fluid having low surface tension in liquid form wets the heat pipe wick, and the liquid is heated and vaporized by the heat-producing component(s) in the usual manner. In one embodiment, the working fluid is freon. In another embodiment, used at very high temperatures, the heat-carrying fluid is an electrically conductive liquid, such as liquid mercury, that fills the circulatory loop, and forcedflow circulation occurs in an endless loop by use of magnetohydrodynamic forces applied to the working liquid. These embodiments appear to require that a closed, serpentine path be provided and filled with a liquid in the substrate, in order to allow forced or natural flow of a heat absorbing liquid throughout the substrate. The liquid may undergo an endothermal phase change at the vaporization temperature, or may absorb substrate heat without phase change.
Neidig, et al. disclose use of soldering or similar attachment of a heat sink to a metallized ceramic that serves as substrate for a semiconductor power module, in U.S. Pat. No. 4,727,454. The heat sink may be a heat pipe with a liquid condensation surface that is larger than the thermal communication area connecting the substrate and separate heat sink, and the heat sink is not an integral part of the substrate.
U.S. Pat. No. 4,859,520, issued to Dubuisson, et al., discloses fabrication of a sintered stack of layers of dielectric material to provide an internal system of liquid-carrying ducts in a substrate for heat removal from the substrate. The substrate is formed from a first layer including electrical interconnect traces and a second layer including the internal ducts used for cooling, the two layers being attached to one another by sintering, gluing or the like, and not being interleaved with one another. A fluorine-containing cooling liquid or gas is circulated through the internal ducts for cooling the substrate. The heated fluid returns through an orifice to a cooling reservoir that is separated from the substrate, and the cooling ducts and cooling reservoir form a closed circulatory system.
Sauzade, et al., in U.S. Pat. No. 4,878,152, discloses use of a graphite core, oriented by material compression in one direction, to provide a material with very high lateral thermal conductivity that connects a printed circuit board with a heat sink. Anisotropy in thermal conductivity across the graphite core layer is used to provide the material with a relatively low effective thermal expansion coefficient by siphoning most of the heat in the lateral direction.